776847 - Create a valgrind tool to taint data sent to and received from content processes

Status: NEW

(Core :: General, defect)

Description

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

We would like to have an automated-as-possible way to ensure we're validating data sent to and by content processes.  A big hammer we could use to hit this problem is a valgrind tool. The way it could work is
 - data sendmsg()d from the master process is tainted unless it passes through a VALIDATE() helper to be marked clean.  Uses of tainted data in subprocesses generate valgrind errors.
 - data we recvmsg() from subprocesses is tainted until a similar validation is marked.  Uses of tainted data in the master process generate errors.

The idea is that we have to go back and review the data we're sending/receiving to make the flood of errors go away.  I fear, however, that VALIDATE()-to-make-valgrind-shut-up would become an anti-pattern.

We'll see by experience how useful this kind of tool would be.